Rights group Amnesty International USA could soon use data analytics to predict which incidents are likely to escalate into larger human rights violations.
As part of an event hosted by New York-based DataKind, an organization connecting volunteer data scientists to nonprofits, about 30 volunteers helped Amnesty create a preliminary program that the global group could eventually use to identify high risk situations.
The program, which volunteers built over the course of a weekend, combed through Amnesty’s 30-year log of “urgent action” calls, or incidents the organization predicts could result in greater human rights violations without intervention. It searched for cases resulting in deaths by execution, and identified terms in the urgent action calls most often associated with executions — assault, attempt, clemency, fear and appeal, for instance. Using the linguistic algorithm, the program was often able to accurately predict which incidents escalated, according to Samir Goswami, managing director of Amnesty International USA’s Individuals and Communities at Risk program.
“If successful, this endeavor may enable those concerned about human rights to more effectively address situations before they reach crisis points,” Goswami said in a statement. The volunteers sorted through 1.4 million lines and 11,000 data files to create this model.
With more research and development over the next couple of quarters, some version of this predictive model could be deployed throughout the organization, Goswami said in an interview. But it’ll likely require a great investment — not only to refine the model but also to build a visual data dashboard for Amnesty officers, and to build a software platform to collect similar data in the future.
And, he added, “I think we need to have some discussions with academics about how we can ethically test this system.”